#asahi-dev · 2026-08-11 16:06
sveni'm writing a tracer for traffic on ring 0 anyway due track down that weird race where some ports are locked and then xdomain doesn't work at all if i remove enough debug prints in apple.c#asahi-dev · 2026-08-11 16:05
sventhen there's just something wrong / new with the Pro/Max variants#asahi-dev · 2026-08-11 16:05
svenbut good to know this works on M1/M1#asahi-dev · 2026-08-11 16:04
svenaw...#asahi-dev · 2026-08-11 16:01
jannauthe ipv6 link local addresses work, iperf3 reports 7.15 Gbits/sec#asahi-dev · 2026-08-11 15:59
svendamn#asahi-dev · 2026-08-11 15:58
jannauargh!, ping probably doesn not work. both devices have the same link local ipv4 address#asahi-dev · 2026-08-11 15:57
nickchanIs that statement slop or something#asahi-dev · 2026-08-11 15:50
sven"In our tests, 13.6 Gbps is within a realistic range. Although this may seem below the raw capacity, real-world factors (including half-duplex operation) usually limit effective throughput." lol#asahi-dev · 2026-08-11 15:45
svenhah :D#asahi-dev · 2026-08-11 15:45
jannautime to debug why iperf3 doesn't work#asahi-dev · 2026-08-11 15:44
jannauhttps://forum.level1techs.com/t/benchmarking-usb4-performance-on-strix-halo/245299/17#asahi-dev · 2026-08-11 15:44
jannauhttps://gist.github.com/geosp/80fbd39e617b7d1d9421683df4ea224a#asahi-dev · 2026-08-11 15:44
jannauthis seems to be already faster than the benchmarks I found intel/amd usb4_net#asahi-dev · 2026-08-11 15:43
maztiny difference... :D#asahi-dev · 2026-08-11 15:37
svennow that's much better :-)#asahi-dev · 2026-08-11 15:37
jannauerr, I typoed the size resulting in fragmentation. `ping -s 1472 -f IP` is now at ~28 gbps#asahi-dev · 2026-08-11 15:36
sventhere's also that magic 0x186 write next to irq enable which might just be some tweaking to when that irq fires#asahi-dev · 2026-08-11 15:34
sveni accidentally added a two or three level indirect pointer lookup plus jump to the hot path so it's not surprising we don't get the full speed, but 2gbps seems low#asahi-dev · 2026-08-11 15:33
jannauiperf3 doesn't work (seem setwork related), flood ping reports 2gbps throughput#asahi-dev · 2026-08-11 15:26
sveni wonder if there's some way to get to that info from macos when not running under the hv#asahi-dev · 2026-08-11 15:24
sventhat should tell you what macos sets for VSE cable info for the cables that don't work#asahi-dev · 2026-08-11 15:24
sventhat's at least in my bash_history#asahi-dev · 2026-08-11 15:24
sventblog=0xffffffff tblogmode=3 tbdp=0xffffffff tbip=0xffffffff are the relevant bootargs I used to enable all verbose debug spam I think#asahi-dev · 2026-08-11 15:23
svenso then something breaks on the non-base M1 chips#asahi-dev · 2026-08-11 15:23
sven\o/#asahi-dev · 2026-08-11 15:22
jannauyes, ping works#asahi-dev · 2026-08-11 15:12
chaos_princessit didn't for me either#asahi-dev · 2026-08-11 15:08
svennow for the interesting part: does ping work?#asahi-dev · 2026-08-11 15:08
svendunno if it'll show up there, i don't have that on my busybox shell :D#asahi-dev · 2026-08-11 15:07
jannauyes, thunderbolt0 net interface is there. I was expecting to see something in boltctl for host2host#asahi-dev · 2026-08-11 15:06
sven*interface#asahi-dev · 2026-08-11 15:06
svendo you see a thunderbolt0 intefcace?#asahi-dev · 2026-08-11 15:06
sventhat looks good#asahi-dev · 2026-08-11 15:05
svenCONFIG_USB4_NET i think#asahi-dev · 2026-08-11 15:05
jannauhttps://paste.debian.net/hidden/2fdb0e03#asahi-dev · 2026-08-11 15:05
svendo you have USB4NET enabled?#asahi-dev · 2026-08-11 15:05
svenis this host2host? not sure that will even show up in boltctl#asahi-dev · 2026-08-11 15:04
jannaugot further along this time but boltctl list is still empty. I see the othewr device announced in dmesg though#asahi-dev · 2026-08-11 15:02
jannauI think it's mostly my fault by trying to use the HV on 2 targets from the same system#asahi-dev · 2026-08-11 14:55
svenuh.. that's not good#asahi-dev · 2026-08-11 14:55
jannauone of the devices froze#asahi-dev · 2026-08-11 14:54
svenbut let me double check#asahi-dev · 2026-08-11 14:54
sveni *think* that is enough#asahi-dev · 2026-08-11 14:54
svenAppleTypeCPhyUSBInterface-debug=0xffff AppleUSBHostResourcesTypeC-debug=0xffff#asahi-dev · 2026-08-11 14:54
svendebug=0x14e serial=3 apcie=0xfffffffe wdt=-1 AppleSynopsysDRDUSBXHCI-debug=0xffff AppleT8103TypeCPhy-debug=0xffff AppleT8103USB40XHCITypeCPort-debug=0xffff AppleT8103USBXDCI-debug=0xffff AppleTypeCPhy-debug=0xffff AppleTypeCPhyCIOInterface-debug=0xffff AppleTypeCPhyDisplayPortInterface-debug=0xffffffff AppleTypeCPhyInterface-debug=0xffff#asahi-dev · 2026-08-11 14:54
svenit already gets far enough, it should find whatever is connected. is that a device or another host#asahi-dev · 2026-08-11 14:53
jannauany non-obvious kconfig option to enable?#asahi-dev · 2026-08-11 14:53
sveni think that's correct because it gets far enough, but can you check what macos would write? let me find the debug arguments you need to get the print